Suitable GPS for GPSDO OCXO

[metrologist]

I wanted to put a few OCXO's I have into a GPSDO build. I was just randomly searching for pps GPS units and came across NEO-7M. I found on the time-nuts mailing list someone mentioned the jitter of these units was ~5ns and perhaps 20ns max, and then it was mentioned they would not be suitable for GPSDO.

I bought a few Trimble units that have spec'd 95ns jitter with some verbiage about taking 100 measurements with 100 samples over a 20 minute period to get the result. I don't if that's good or not, but those are the units I ordered.

If they are not suitable for GPSDO then I can use them for other less sensitive timing.

The basic idea I had for the GPSDO was to use a micro clocked by the 10MHz OCXO and count clock cycles between the 1pps signal, then do some fancy math over a period of time to slowly drive the control voltage with a ~12 bit dac. I was hoping this would accommodate for any jitter.

I'm sure there are designs already out there, I see one on the instructables site and I briefly caught another using an atTiny.

Any ideas on this or suggestions of where to go looking for more info? I'd need some help probably with the math programming and maybe some of the circuit design and component selection. I'd probably use an Atmega 328p.

[Gyro]

You want a 'timing' receiver rather than a 'location' one. For Ublox that means an LEA-xT rather than a NEO-xM (don't think there's a NEO-xT). For Trimble, a Resolution-T (or -SMT), or whatever the later models are, etc.

It's worth trawling the specs on the manufacturers' sites.

Examples:

http://www.ebay.co.uk/itm/Trimble-Resolution-T-GPS-Timing-Receiver-Module-1PPS-Output-15-nsec-accuracy-/162586564826?hash=item25daea18da:g:OogAAMXQY8JRgFFR

http://www.ebay.co.uk/itm/Trimble-GPS-SMD-IC-Multi-GNSS-Timing-Module-Resolution-SMT-GG-P-N-89999-00-/262621138349?hash=item3d25708dad:g:cdIAAOSwNphWWqGt

https://www.ebay.co.uk/sch/i.html?_odkw=LEA-6T&_sop=15&_osacat=0&_from=R40&_trksid=p2045573.m570.l1313.TR0.TRC0.H0.TRS0&_nkw=LEA-6T&_sacat=0


Edit: Note that the Resolution-SMT will be a 'challenge' to breadboard with.

[MosherIV]

The basic idea I had for the GPSDO was to use a micro clocked by the 10MHz OCXO and count clock cycles between the 1pps signal, then do some fancy math over a period of time to slowly drive the control voltage with a ~12 bit dac. I was hoping this would accommodate for any jitter.

I'm sure there are designs already out there, I see one on the instructables site and I briefly caught another using an atTiny.

Any ideas on this or suggestions of where to go looking for more info? I'd need some help probably with the math programming and maybe some of the circuit design and component selection. I'd probably use an Atmega 328p.

I have been thinking of doing it the same way.

I do not think you need to worry about the maths, the instructables way is simple enough - it just averages the count over a very long period (1, 2, 4, 12, and 24 hours). This is enough to average out the jitter in the PPS. Since you sound like you are using a comercial OCXO, their drift is usually less than 20ppm/year so this technic should be fine.
Basically, the count should be dead on 10MHz (+/- 1 count because of the PPS jitter - you cannot over come this unless you sychronise the 10MHz to the PPS but then you will end up adding jitter to the 10MHz, so in my opinion it is better to live with the 1 count error and have the stability of the OCXO) if the count is off, then you adjust the 10MHz. Whether you do this on the 1, 2, 4, 12 or 24hour is up to you.

Do not forget to have some non volatile storage to save the current DAC output, for cold starts and when GPS lock is lost.
Obviously, do not average when GPS lock is lost.

As Gyro pointed out the M variant of the Ublox devices are the general purpose positioning devices. The T variant is the timing variant but just like you, I found they are more difficult to get hold of.
The main difference from your point of view is the M variant does not have a TCXO so the temperature of the GPS module will have an effect on the stability of the PPS.
Other differences are that it has less facility for timing eg does not provide 10KHz signal. (other GPSDO seem to be beased on PLL the 10MHz to the 10KHz because the 10KHz is derived off the timing from the satellite signals).

I am curious, you say you already have a Trimble GPS. There are other project blogs on the web which use the Trimble. The Trimble has the 10KHz signal.

I will follow your progress with interrest. Good luck and have fun 

[Gyro]

I'm not sure if you saw the ebay links I added to my reply. It's the first time in a while that I've seen a Resolution-T listed (note that it only does 1pps though).

The LEA-6Ts are still reasonably available in those Chinese "Ublox LEA-6T GPS Module w/Compass for APM2.5 APM2.6 Flight Controller Multirotor" modules that I did a teardown on.

It depends on how fine the tuning range of the OCXO is of course but I have a gut feel that 12 bit resolution might not be enough.

[MosherIV]

Thanks Gyro, did know you could get those.

Just quickly read your tear down 

Given the choice.....go for the T variant. It is less affected by temp of the module.

Is there a way to seperate the areial and add a smb connector for external ariel?

[Gyro]

You're welcome. Yes, definitely go for the T version, it's not just the TCXO, the internal firmware is also optimized for timing duty. The LEA also gives you two programmable frequency outputs versus one in the NEO.

Well it's easy to strip the existing antenna, LNA and related components and substitute a coax, which could have a bulkhead mount SMB on the end. I did it out of necessity when I 'lost' the LNA ( ). The whole saga is documented on page 2 of the teardown thread... https://www.eevblog.com/forum/projects/ebay-u-blox-lea-6t-gps-module-teardown-and-initial-test/ for reference.

Here's what it looked like, ignore the little inductor - that's only there because I killed the internal one! The consolation is that it works better with a cheap external active antenna than it did with the existing patch one. It also keeps the frequency outputs inside the GPSDO box rather than on the far end of a cable.

[suicidaleggroll]

Basically, the count should be dead on 10MHz (+/- 1 count because of the PPS jitter - you cannot over come this unless you sychronise the 10MHz to the PPS but then you will end up adding jitter to the 10MHz,

No need for synchronization if you code it right.  The frequency error will accumulate until you get to an extra count.  Eg: if the OCXO is putting out 10,000,000.1 Hz, you'll get one extra count every 10 seconds, which will go into the averaging and affect the tuning.

[metrologist]

I already bought one of the Trimble based GPSDO systems and wanted to build a few more of my own since I have several good quality OCXO's. I'm not sure if the GPSDO I bought is an upgraded version of the ThunderBolt (newer gen?), or probably just something built for the telecom industry. It is the system on ebay that is similar to the BG7TBL, with the sine and square wave 10 MHz and 1PPS outputs.

I bought a couple of these GPS modules just because they were $3 (just the GPS module): ftp://ftp.trimble.com/pub/sct/embedded/bin/Manuals/Lassen_LP.pdf

I figured I could average out the jitter and only slowly steer the OCXO with the micro to keep the long term drift in line, maybe just run a calculation a couple or few times per day and slowly adjust the dac between those intervals. I do not know how the commercial GPSDO's control their oscillators. I'm thinking that would be a better solution than trying a PLL and introducing jitter.

If I understand how to apply my 10 MHz OCXO spec, it is 5E-10/day aging. I think that is 5 mHz/day, or about 2 Hz/year? After typing that, it seems that the micro would not be able to count an offset until it saw one cycle out, which would take half a year? Maybe I'd have to use a phase detector and apply the same approach. Is that even possible? It would seem a better than 0.5ns resolution phase detector would be needed to even see something after a day.

I have so much learn and relearn about this.

[suicidaleggroll]

If I understand how to apply my 10 MHz OCXO spec, it is 5E-10/day aging. I think that is 5 mHz/day, or about 2 Hz/year? After typing that, it seems that the micro would not be able to count an offset until it saw one cycle out, which would take half a year? Maybe I'd have to use a phase detector and apply the same approach. Is that even possible? It would seem a better than 0.5ns resolution phase detector would be needed to even see something after a day.

5e-10/day aging is 5 mHz/day.  So after 1 day, the frequency will be off by 5 mHz.  At a 5 mHz frequency offset, it would take 200 seconds to accumulate an extra cycle.

I think you're worried about the wrong thing though, aging isn't really the problem, it's frequency stability over temperature.  Most OCXOs only have a temperature stability of around 1e-7 over their operational temperature range.  Unless you're putting this thing in a VERY tightly controlled environment, things like A/C or a heater kicking on and changing the room temperature by 1-2 deg C is going to have a much larger effect than the per day aging.  Your MCU needs to be able to react and tune the OCXO much faster than a few times a day, I'd say once a minute is probably the slowest you'll want to do it.  I can't speak for all commercial GPSDOs, but the few I've tested update every second.

[metrologist]

5e-10/day aging is 5 mHz/day.  So after 1 day, the frequency will be off by 5 mHz.  At a 5 mHz frequency offset, it would take 200 seconds to accumulate an extra cycle.

OK, that should have been obvious... I think that would even need to be integrated over time (where are my calculus books?).

The temp spec is 2E-10, 0C to 55C. Would it help to have it in a well insulated box?

Short term stability is 5E-12/s

Now I am more motivated again 

I think I would want 100 averages at least to deal with this GPS's 95ns jitter.

[metrologist]

He took my offer for two. He has about 70 units.

http://www.ebay.com/itm/Trimble-39263-Lassen-Low-Power-LP-GPS-Interface-Module-Timing-1PPS-MCX-M-SMA-F-/401314512604

[MosherIV]

Most OCXOs only have a temperature stability of around 1e-7 over their operational temperature range.  Unless you're putting this thing in a VERY tightly controlled environment, things like A/C or a heater kicking on and changing the room temperature by 1-2 deg C is going to have a much larger effect than the per day aging.  Your MCU needs to be able to react and tune the OCXO much faster than a few times a day, I'd say once a minute is probably the slowest you'll want to do it.

I would hope that a ovened crystal is better than that ! Otherwise why would they have bothered to build a crystal into an oven !

Typically the oven temp is 45 to 50'C, well above ambient so that ambient should not affect it.

If the OCXO had to be adjusted every minute, the GPSDO would not be that stable.

[CJay]

He took my offer for two. He has about 70 units.

http://www.ebay.com/itm/Trimble-39263-Lassen-Low-Power-LP-GPS-Interface-Module-Timing-1PPS-MCX-M-SMA-F-/401314512604

Damn shame he won't ship outside the US, that's a *really* good price

[edpalmer42]

Most OCXOs only have a temperature stability of around 1e-7 over their operational temperature range.  Unless you're putting this thing in a VERY tightly controlled environment, things like A/C or a heater kicking on and changing the room temperature by 1-2 deg C is going to have a much larger effect than the per day aging.  Your MCU needs to be able to react and tune the OCXO much faster than a few times a day, I'd say once a minute is probably the slowest you'll want to do it.

I would hope that a ovened crystal is better than that ! Otherwise why would they have bothered to build a crystal into an oven !

Typically the oven temp is 45 to 50'C, well above ambient so that ambient should not affect it.

If the OCXO had to be adjusted every minute, the GPSDO would not be that stable.

Although a temperature of 45 to 50C is possible, it would be unusually low.  A more typical range would be 60 to 80C for both AT and SC cut crystals.  The idea is to set the temperature to the point where the change of frequency with temperature is zero.  For optimum performance, each oven has to be adjusted seperately because each crystal is unique.

Ed

[suicidaleggroll]

I would hope that a ovened crystal is better than that ! Otherwise why would they have bothered to build a crystal into an oven !

Because without the oven it would be even worse, by several orders of magnitude.

[metrologist]

Those receivers are on sale today, less than I paid. Now I paid too much! I will buy a couple more when I get home if they are still on sale...

[CJay]

Those receivers are on sale today, less than I paid. Now I paid too much! I will buy a couple more when I get home if they are still on sale...

Well if you fancy buying a couple for me and shipping them...

[metrologist]

I've never shipped anything out of the country. I did just buy 3 more (he wasn't as nice on the shipping this time, but for $2.19 I wasn't going to haggle on the price) and could see how they turn out. I really do not need 5 of them 

I wonder if they have a chance at receiving using a passive antenna? I wanted to try building a quadrifilar helical antenna after my failure at a simple helix wire wrapped around a plastic form with a disc ground plane.

These plans for a stationary install, and I think I'd need a 1nF DC blocking cap:
http://www.i1wqrlinkradio.com/antype/ch30/chiave1743.htm

If it does not work then I'd just buy a few more Chinese active puck antennas...

[cdev]

A turnstile antenna would be easy to make and tiny. Also you could make a "pinwheel" type.

[metrologist]

I'm also looking at this online calculator, but I do not understand the total compensated lengths.

http://www.jcoppens.com/ant/qfh/calc.en.php


I don't even understand the first example I posted because if you add up each segment, the length does not come out...

For example, in the first reference, I will add 30 + 55 + pi15 = 132mm (and that is not accounting for the small radius at the bends), but the length specified is only 105mm.

The online calculator is giving me a corrected length of 211.8 and 201.5 with approximate diameter of 26.7. It's longer and narrower (I used 0.4 ratio), but still does not add up to anything close.

I had good luck building a 23cm band bi-quad reflector with lips and successfully used it with a C-band TV satellite dish in place of the original horn LNAs. I measured ~10dB gain relative to a dipole just for the bi-quad reflector. I also tried a double bi-quad for the 2m band. It was fun. We'll see how it goes, maybe tonight if I find some time and materials laying around.

[cdev]

A small QFH would be the best but not so easy to get right if you're making it out of wire. Copper tape around a low loss form of exactly the right size would be better. Also you'll likely need to feed it using semi rigid coax.

Take a look at the ones used in radiosondes.

The passive ceramic patch antennas from Taoglas work quite well and aren't that expensive.

Only a small subset of GPSs are really suitable for use in a GPSDO.

[cdev]

I have a cheap timing antenna and it is a QFH. I paid around $20 for it on ebay. Its the one that looks a bit like a soft freeze ice cream cone. Sold under Lucent, MaxRad and a few others.

Quote from: evb149 on Today at 14:00:00

When it comes to antennas, size and pattern matters.  So from that standpoint yeah building (or buying) a "big" (or "bigger" anyway) good antenna will outperform a little puck patch assuming you give it also whatever LNA / filter it may need.

The trick is just building an antenna that actually works well without any tuning equipment besides the "bend the coat hanger" approach.  Even just replicating a published design with slightly different materials and construction could alter performance enough to disappoint.

I'd like to try a QFH or some of the other antennas also to improve my timing signal reception.  But haven't yet tried nor bought any equipment needed to quantify the actual performance.

For precision timing stuff I've often heard to use "survey / timing grade" antennas with good hemispheric coverage to the horizon but also a design / mounting that suppresses multi-path problems which can be a big problem assuming you passed the "pretty clear view of the sky" hurdle.

Hard to swallow looking at $xxx or $xx antennas for a $3 receiver, though, so, yeah, building.....maybe.

Let us know if you come up with a great and reproducible design!

[metrologist]

I'll give it a wing. I can also get a smith plot at work as we have a few VNA's.

I do have a commercial helix, about the size of my thumb with a recessed SMA male that will screw right on to that pigtail, as long as I remember to decouple the DC. I'll have to make up an adapter.

[edpalmer42]

There are a few things to keep in mind about antennas.

Most timing receivers monitor the current fed to the antenna and get upset if the current is too low or too high.  If you build your own antenna, you'll have to include a fake load.  Check the data sheet for your receiver.

If your receiver is a bit older, it may be a bit deaf.  Without the gain provided by an active antenna it might have trouble.

Timing grade antennas don't have to be expensive.  Looks like they start in the $20 - $30 range.

If you buy an antenna, make sure that the voltage the GPS receiver puts on the antenna lead is compatible with the antenna.  Really old or weird receivers used +12V.  Most older ones use +5V.  The newer ones use 3V3.  Lots of antennas are +3V3 and +5V compatible.

Ed

[cdev]

Is your helix a Sarantel antenna?  They do a size reduced GPS QFH thats pretty cool.

A QFH is a very good all around antenna with a uniquely omnidirectional pattern that is not effected practically at all by its surroundings, and maintains the right hand circular polarization, nomatter what direction the signal is going to, coming from. It may actually be the best antenna if you are not using multiple constellations.  (more than just GPS L1) - the problem is a QFH will work within its own band, and likely also has a built in SAW filter so isnt so great for signals from other GNSS systems than its designed for.

A QFH is certainly an antenna that will do excellently for GPS timing.


Quote from: metrologist on Today at 15:01:17

I'll give it a wing. I can also get a smith plot at work as we have a few VNA's.

I do have a commercial helix, about the size of my thumb with a recessed SMA male that will screw right on to that pigtail, as long as I remember to decouple the DC. I'll have to make up an adapter.